Hi!

On 2019-09-06T09:02:13-0700, Julian Brown <julian@codesourcery.com> wrote:
> This patch adds profiling support to the AMD GCN libgomp plugin, modeled
> after the equivalent support in the NVPTX plugin. This gives a positive
> test delta in AMD GCN offload testing.

Yay!  \o/

> I will apply to the openacc-gcc-9-branch shortly.

..., and later these changes got into master branch, via integration into
"[PATCH 7/7 libgomp,amdgcn] GCN Libgomp Plugin".

> --- a/libgomp/plugin/plugin-gcn.c
> +++ b/libgomp/plugin/plugin-gcn.c

|  static void
|  gomp_offload_free (void *ptr)
|  {
|    GCN_DEBUG ("Async thread ?:?: Freeing %p\n", ptr);
|    GOMP_OFFLOAD_free (0, ptr);
|  }

> @@ -3046,6 +3075,35 @@ GOMP_OFFLOAD_free (int device, void *ptr)
>        return false;
>      }
>
> +  struct goacc_thread *thr = GOMP_PLUGIN_goacc_thread ();
> +  bool profiling_dispatch_p
> +    = __builtin_expect (thr != NULL && thr->prof_info != NULL, false);
> +  if (profiling_dispatch_p)
> +    {
> +      [...]
> +      prof_info->event_type = acc_ev_free;
> +
> +      [...]
> +      GOMP_PLUGIN_goacc_profiling_dispatch (prof_info, &data_event_info,
> +                                         api_info);
> +    }
> +
>    return true;
>  }
>

> @@ -3276,6 +3334,35 @@ gcn_exec (struct kernel_info *kernel, size_t mapnum, void **hostaddrs,
>       {1,       64, 16}
>      };
>
> +  struct goacc_thread *thr = GOMP_PLUGIN_goacc_thread ();
> +  acc_prof_info *prof_info = thr->prof_info;
> +  acc_event_info enqueue_launch_event_info;
> +  acc_api_info *api_info = thr->api_info;
> +  bool profiling_dispatch_p = __builtin_expect (prof_info != NULL, false);
> +  if (profiling_dispatch_p)
> +    {
> +      prof_info->event_type = acc_ev_enqueue_launch_start;
> +
> +      [...]
> +      GOMP_PLUGIN_goacc_profiling_dispatch (prof_info,
> +     &enqueue_launch_event_info, api_info);
> +    }
> +
>    if (!async)
>      {
>        run_kernel (kernel, ind_da, &kla, NULL, false);
|        gomp_offload_free (ind_da);
|      }
|    else
|      {
|        queue_push_launch (aq, kernel, ind_da, &kla);
|        if (DEBUG_QUEUES)
|       GCN_DEBUG ("queue_push_callback %d:%d gomp_offload_free, %p\n",
>                  aq->agent->device_id, aq->id, ind_da);
>        queue_push_callback (aq, gomp_offload_free, ind_da);
>      }
> +
> +  if (profiling_dispatch_p)
> +    {
> +      prof_info->event_type = acc_ev_enqueue_launch_end;
> +      enqueue_launch_event_info.launch_event.event_type = prof_info->event_type;
> +      GOMP_PLUGIN_goacc_profiling_dispatch (prof_info,
> +                                         &enqueue_launch_event_info,
> +                                         api_info);
> +    }
>  }

Per that, we've currently got:

  - [...]
  - acc_ev_enqueue_launch_start
  - launch kernel
  - free memory
  - acc_ev_free
  - acc_ev_enqueue_launch_end

This confused another thing that I'm working on, so I adjusted that to:

  - [...]
  - acc_ev_enqueue_launch_start
  - launch kernel
  - acc_ev_enqueue_launch_end
  - free memory
  - acc_ev_free

Pushed to master branch commit 649f1939baf11f45fd3579b8b9601c7840a097b3
"Fix OpenACC/GCN 'acc_ev_enqueue_launch_end' position", see attached.


Grüße
 Thomas


-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955